Atrial fibrillation (AF) is a common arrhythmia, affecting almost 1% of the US population. Left atrial (LA) fibrosis has been identified in patients with AF, using biopsies, post-mortem analysis, and electrophysiological testing. However, the role of fibrosis is not known-whether it is a precursor to AF, maintains AF, or is a result of AF. Cardiovascular magnetic resonance (CMR) is capable of identifying scar with unparalleled precision, using the late gadolinium enhancement (LGE) technique. However, the thin myocardial LA wall presents challenges to identifying scar, especially for patients not in sinus rhythm. This proposal aims to improve the LGE method for detecting pre-existent scar. Firstly, the spatial resolution of LGE CMR method will be increased using increased SNR from a 32-element phased array cardiac coil. Image edge- enhancement will be reduced in the LGE images, using novel pulse sequence modifications to maintain necessary fat-suppression. A time-study of scar enhancement after injection of contrast agent will discriminate between scar and image artifacts. Many AF patients are not in sinus-rhythm. We propose a method for reducing artifacts due to RR variability. Using this LGE CMR method, we will study control subjects and patients prior to undergoing treatment either by catheter ablation, minimally invasive surgery, or drug therapy. The scar detected prior to therapy will be correlated with clinical outcomes. It will also be correlated with other methods for detecting scar--electrical mapping of the left atrium in patients undergoing catheter ablation, and histological analysis in patients undergoing surgery. This proposal will establish LGE CMR as method for detecting pre- existent fibrosis in AF patients.